🐛 FIX: SQLite: apply escape \ in QueryBuilder like and ilike

[zhubonan]

For like and ilike, the special characters _ and % may need to be escaped if they have literal meanings (LIKE .. ESCAPE in SQL) . However, the sqlite backends do not implicity treat \ as escape sequence (unlike Postgres). https://sqlite.org/lang_expr.html

I discovered this bug when using the SQLite backend (in memory) with aiida-vasp where one of the entry point has name with an underscore. Querying such node gets escaped like:

QueryBuilder(path=[{'entity_type': 'data.vasp.potcar_file.PotcarFileData.', 'orm_base': 'node', 'tag': 'PotcarFileData_1', 'joining_keyword': None, 'joining_value': None, 'edge_tag': None, 'outerjoin': False}], filters={'PotcarFileData_1': {'node_type': {'like': 'data.vasp.potcar\\_file.%'}}}, project={'PotcarFileData_1': []}, order_by=[], limit=None, offset=None, distinct=False)

Here the _ is automatically escaped, but the query only works fine with the psql backend, with sqlite it does not match any existing objects.

This PR makes \ to pass explicity as escaping character for sqlite backends. The use of \ for escapign is already hard-coded aiida.

Here is an example test script:

from aiida_vasp.data.potcar import PotcarFileData
from aiida import orm


from aiida import load_profile, engine, orm, plugins
from aiida.storage.sqlite_temp import SqliteTempBackend

profile = load_profile(
    SqliteTempBackend.create_profile(
        'myprofile',
        sandbox_path='_sandbox',
        options={
            'warnings.development_version': False,
            'runner.poll.interval': 1
        },
        debug=False
    ),
    allow_switch=True
)

d = PotcarFileData()
d._store()
q = orm.QueryBuilder()

# Query the PoatcarData
q.append(PotcarFileData)
print(q)
# Should get one node
print(q.all())

d1 = orm.Data()
d1.set_attribute('a', 'a_b')
print('d1', ' ', d1.store())

d2 = orm.Data()
d2.set_attribute('a', 'acb')
print('d2', ' ',d2.store())

d3 = orm.Data()
d3.set_attribute('a', 'a\_b')
print('d3', ' ',d3.store())

q = orm.QueryBuilder()
q.append(orm.Data, filters={'attributes.a': {'like': 'a\\_b'}})
# Should match to d1
print(q.all())


q = orm.QueryBuilder()
q.append(orm.Data, filters={'attributes.a': {'like': 'a_b'}})
# Should match to d1 and d2 but not d3
print(q.all())

q = orm.QueryBuilder()
q.append(orm.Data, filters={'attributes.a': {'like': 'a\\\\_b'}})
# Should match to d3
print(q.all())

p1 = orm.WorkflowNode()
p1.set_process_type('aiida.test.a_b')
print(p1.store())
p2 = orm.WorkflowNode()
p2.set_process_type('aiida.test.abb')
print(p2.store())

q = orm.QueryBuilder()
q.append(orm.WorkflowNode, filters={'process_type': {'like': 'aiida.test.a_b%'}})
# Should match to p1 and p2
print(q.all())

q = orm.QueryBuilder()
q.append(orm.WorkflowNode, filters={'process_type': {'like': 'aiida.test.a\\_b%'}})
# Should match to p1 only
print(q.all())

[chrisjsewell]

Yep looks good cheers